#πο½general-with-images
1 messages Β· Page 125 of 1
A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type.
gg?)
final gif.. sketchey but i'll run with it
Whats your GPU? You may need to add --no-half-vae to the webui-user.bat
Please draw a picture with a dark blue background to represent technology and digital intelligence. The picture reflects elements such as artificial intelligence, automation, factory, robotic arm, code, etc. Thank you.
Here is the image you requested.
Bro. Stop taking pictures in my room.
Please draw a picture with a dark blue background. The picture size is 16:9, showing technology and intelligence. All elements are blue.
Here is the image you requested.
Please draw a picture. The ratio of the picture is 16:9, and the main color of the picture is dark blue. The theme of the picture is technology and intelligence. Thanks.
Here is the image you requested.
Please draw a picture. The ratio of the picture is 16:9, and the main color of the picture is dark blue. The theme of the picture is technology and intelligence. There is a robot with no face, teeth or mouth in the picture. Thanks.
what else ya got? I'm testing out a regional prompter automation.
hah you don't know sdxl very well if you're asking for realistic fingers. π
OK
Here is the image you requested.
hah very cool
cool
very neat
Here is the image you requested.
very impressive
@nimble mason very good!
Beep boop! I am pleased! Do you have another request!
@nimble mason please give me another picture, the same content,but the style different.thank you.
my automation thrives on uniqueness. uniqueness quotient unsatisfied.
here's a robot instead.
Here is the image you requested.
give me another idea. doesn't matter what it is.
should have more than one subject in the picture
good job.
Please draw a realistic picture of a robot with dark blue as the main color. The pictures reflect artificial intelligence, code and other elements, thank you.
Here is the image you requested.
Please draw a realistic picture of a robot with a ratio of 16:9 and dark blue as the main color. The pictures reflect artificial intelligence, code and other elements, thank you.
An electromagnetic gun made of a plastic pipe with enameled wire wrapped around it.
An electromagnetic gun made of a plastic pipe with enameled wire wrapped around it.
An electromagnetic gun made of a plastic pipe with enameled wire wrapped around it. @nimble mason
Here is the image you requested.
thats wicked
Here is the image you requested.
zombie dance party.
can we drop in gif?
yep
wicked
lemme make sure this thing barfs out something good
playing with SVD
cat needs his nip
twitchy little bugger
haha very good
changing the fps back down
rendering 90 video frame / motion of 20 / 24 fps on the condition.
wondering if the condition was fine
video combine is now at 24 fps
ohgod this next one is wierd
1050ti
i have it
wonderful!
create a big pig use white color,thank you.
yummy.
now I wanna see a human as chocolate
thank you.
/img textbook cover titiled "Principles of Marine Sediment Dynamics"
good job.
this image what you request
thx
@clever oar Are you a robot ?
Please draw a picture. The ratio of the picture is 16:9, and the main color of the picture is dark blue. The theme of the picture is technology and intelligence. Thanks.
Please read #1047610792226340935 . Thanks
a girl sitting on the beach
#1047610792226340935 not bots online
is there a place where people share their generated pictures with prompts and settings so we can learn from them? i know prompthero. would like more suggestions. xx
Civitai
aha. thank you!
@dense tapir Are you here at the moment?
Sharkvacado
π
Ideogram
#πο½general-with-images cat with funny nose
#1047610792226340935 Bots currently offline. You can run Stable Diffusion locally or use online services.
An experienced warrior with a large magical sword in 3D, showing his full body, and with a white background and his arms at the sides of his body.
cat with funny nose
#1047610792226340935 Bots currently offline. You can run Stable Diffusion locally or use online services.
sharks sued for using their image and discrediting
Here is the image you requested.
Here is the image you requested.
"Design a minimalist logo for the brand 'ζΌ«δΊθ³' (Manting Fang). The company deals with agarwood, incense, and related products, and they want the logo to include elements related to fragrance."
Here is the image you requested.
Reminds me of nibbler in futurama with his "dark matter"
cascade is better then xl?
Cascade is better than sdxl if it had the fine tunes that sdxl does. Seeing that it probably never will, it's not better, just different.
Proteus beats cascade
My cascade workflow does cascade and then uses the latest proteus as a refiner. Cleans up faces etc.
π€
You use in comfy UI right
Yeah, I could never get cascade working in a1111
What model is that
That's prolly a multi step composition
snakesberry
Is there any img2vid workflow for animatediff available?
Was this trained on egirls too? π
yes
deadass
i trrained it off discord egirls π₯
now im testing training data output
success
could be very illegal i aint ask for permission
π
I'm unsure about the legality because the pictures were presumably posted publicly
But I won't dive into that
vs cascade
who did it better @gusty cloak
cascade or the other
prompt: 18 YEAR OLD JAPANESE GIRL!! WEARING BUNNY EARS PINK OUTFIT FLUFFY!!!! WEARING CHOKER EATING ICECREAM!!
!!!
ye the first one beat cascades ass
tell me something
to gen
I don't know π
Everything that comes to mind is "Nope I'm not making this person generate that" (it's completely innocent just awkward π)
tell me
i wont find it awkward
99% of the time I do male portaits π€
dont mind
It's just that 99% of Civitai for example is "Whamen!!!"
just send
π
prompt
Alright let me think of one
ping me for each prompt
@sturdy wraith
A young man with short hair and a small beard, wearing a black hoodie, holding a sign that says "BOTS DEAD", looking at the camera and smiling
(Didn't specify hair color this time just to see what your model gives us!)
He looks a bit like a child π Too many egirls in the training data π
wait
i just realisd
yk those adult servers
that ask you to hold a paper with text on it
to confirm youre that old
what if you used this ai π
π€£
let me try
what is egirls emo girls help training ai?
yes
they make it easier
yk how egirls write text
on paper
and show it
on discord atleast
and write shit on their arms and body
it helps training because then you can get proper text output
why is not do boys
I would not use a Discord server for dating π
only girls
You're asking the real questions π
trying to understand the right steps on this workflow
lol my english not good
What I said meant that you asked a good question π
DEβD DED
lol
when i finetuned oin the egirl data
now the ai zero shots text
especiall if its on paper
example
thats odd
how egirls works you go to channel and coolect picture with text?
lol
imagine this model with T5
what model
I HAVE.... to be doing something wrong here
how
I cant get a sharp image
dont undertand all process
my images arent sharp either
skorchekd
β
15/03/2024 22:24
https://youtu.be/70H03cv57-o?si=LNTsG3BEdL6lOGtq r
LORA is a fantastic and pretty recent way of training a subject using your own images for stable diffusion. Say goodbye to expensive VRAM requirements and hello to this innovative new way of fine-tuning! In this video I will show you how to train a LORA weight using the kohya ss GUI with less than 7GB of VRAM and how you can then use those LORA ...
follow this
although
i dont use low vram config
i use high vram
head of dead
perfect
dead of head
head with bed)
note perfect
look my result )
thats coldc
"Radiating with an eerie blue glow, Godzilla towers over the cityscape, its colossal form a silhouette against the night sky. From its gaping maw, a torrent of green fire erupts, but this is no ordinary blaze β it's infused with the searing energy of radiation, twisting and contorting with an otherworldly intensity.
The flames cascade downward in a mesmerizing display, engulfing buildings in their azure embrace. Each structure consumed by the inferno seems to pulse with a vibrant, unnatural light, as if the very atoms within are being torn apart by the overwhelming power of Godzilla's breath.
The air crackles with the sheer energy unleashed, sending shockwaves rippling through the surroundings. The ground quakes beneath the weight of the behemoth as it strides forward, its steps leaving behind trails of smoldering devastation.
In this moment of chaos, Godzilla embodies a primal force, an avatar of destruction fueled by the madness of a world gone awry. Its radiating blue scales gleam with an ominous brilliance, casting an eerie glow upon the ruined city below. This is not just a creature of flesh and bone β it is a living embodiment of insanity, a relentless force of nature that knows no bounds."
prompt following insajneee
So I am having issues with the Stable DIffusion WebUI.
For starters here are my specs:
Gpu: Nvidia RTX 3060
Cpu: Intel i7 (13th gen I believe)
And for some reason no matter the steps I have for an image to be generated, and no matter the size or sampling method, images are generated in less than five seconds, whereas with SDXL it takes about a minute.
My settings are in the picture and the images generated all have poor quality.
My 3069 takes 30 seconds to gen sdxl
Here's the image you requested
Jr using a easy to run model that's why 5 seconds
Should I uninstall Stable DIffusion and reinstall it. Because all the images are coming out poor quality for some reason
hello
Bro
Use upscale
Res fiz
Fix
But will it fix the ugliness of the image?
Aight, let me try
Then send the output here
Don't fuck with the settings
Hit generate and send image here
How about this one
best
SD is fucking amazing
my friend says that creating an interesting image is entertainment for 2 times, but I say it can be done endlessly
we have different mindsets
he dont undestand it
Yeah the dopamine we get from generating images is stupid
no
still ugly, and still generates the same rate at 150 steps
I might have to uninstall
Challenge: To go back to the good old days, use base SD 1.5 to generate something π
Here's the image you requested
i close to this challenge)
Challenge accepted
That's 1.5
Damn! I don't buy it π
all my models on 1.5 base
Lol but i am using a websiteπ
Not a 1.5 checkpoint though, just the base SD 1.5
You know, the thing that generates our beloved garbage portraits!
I think it is the base model
Idk maybe it's the schedule im using
Yeah it's definitely 1.5
DPM++SDE and upscaling doing the heavy lifting, but that's clearly 1.5 base
I wonder why so many images in 1.5 are out of bounds π€
My bro
Why are u using 120 steps
π
The model Ur using is prob ass
What happened, the picture above looks way better but it's the same model, maybe this website is fucking with me
SD 2.1
CFG?
Beats me, 2.1 for ya
24 / 8
Post Malone deepfake
I stopped caring that my computer was flirting with me last nightβ¦ just couldnβt get animations to work right. lol
Normal 7.5, I'm using a websiteπ
try 4
I prompted for van Gogh that's why
AI porn.. ugh. "just fine" its not. its ugly. Real performers are superior. Just use that for your needs.
Steps are image clarity, and CFG is nightmare fuel right now?
2.1 pretty good at photorealism
Right* >?
i remember why i don't bother coming into the images channel too often
Itβs wierd here
Just no egirls plz
yeah that's what I used it for back in the day π΄
I was kinda ignoring the prompts and working on the rest of the junk, used Godzilla for the rest
a non josh hartnet version of post malone, brought to you by instant-id
The race isn't the same
He looks Italian
And the tattoo and beard are wrong
Yeah
diskord eboys
malones beard is ever changing. yeah face swaps can't do face tattoos . you gotta prompt for them. case in point? your attempt.
whats with racism against italians? what is this early 20th century new york? post malone is lliterally italian descent too
malone may have been a clue
Sushi-zilla
delicious
What π
Racism against Italians?
Racism only exists because people pull the race card too much
Wild
I didn't even have a racist thought in my mind I deadass just thought the image appeared Italian
oh. you're one of those anti woke people. noted.
I'm not anti anything I'm neutral
lol right after this you proceed to say "i'm neutral" >.... i bet you actually believe that. Cognitive dissonance. What a bitch.
π€
Who let Godzilla in the kitchen
Juggernautv9 but not the lightning version
where my image what i request)
Prompt?
sonic the hedgehog made of glass
also your xl models listen promt better than my 1,5
before i think xl only for realism and bad for details
hand discrimination bad hands deserve a photo too
deformed hands metter)
Could be better with a different prompt
I was asleep
How are you running SD, locally?
yes
wow nice
A1111?
yes
How about you check out this vid
I am back from a break from the channel! After taking the QA job at Stability (the company that makes stable diffusion) I needed to balance life and work, so needed to take a break. However, now that convention season is pretty much at an end, I can get things back on the road! This week I wanted to cover some AI prompt engineering tricks I d...
To test if it generated just like that. And the model I used was Anything V3
Did you make make use of those syntax, try something like "photo of (a tick:0.5), photo of (a spider:0.5) then adjust the weights according, its always good to use a fixed seed if you're a beginner at using these syntax
give me something
to generate
anything you want
A black haired demon girl wielding a holy sword
Maybe a tick and a turtle, tick and spider is already similar
@lapis quarry yo if you need help let me knpw
i got the same gpu as you
i can give you some of the loras i use etc etc
π
: D
""Cat girl people losing minds
lol
Here's that tick-turtle
π
Try throwing a glowing bioluminescent jellyfish in there somewhere
1 sec
A.I. always fun π
Macro photo futuristic glowing bioluminescent [tickπͺΌ10] made of glass, underwater, dramatic light, ultra realistic, 32k resolution, national geographic cover photo, lora:add-detail-xl:1 lora:detailed_notrigger:1
man, are you using comfy?
"Mech warrior cat girl zombie horde"
Stable Diffusion Forge
You can get the same results in comfy, it's actually easier there
the model im using is very good for prompt following ngl
it does whatever i put
I'm using Forge too. Did you use ControlNets to achieve these results?
No ... prompting and 2 Loras only ...
Watch the YouTube video i linked above
Good base model then
hephaistosNextgenDPO_v10.safetensors
checking it
thanks
Ooo send em over
Many good models these days someone here told me about it ...
how much vram
is you 3060
how the fuck do i send
my brain
prompt?
Already doing next one, sorry ...
Kinda: (Double exposure effect:1.2) ruins of mysterious abandoned post apocalyptic (US Capitol:1.4) zombie body portrait silhouette, superimposed by Conrad Roset, Brandon Kidwell, Andreas Lie, Dan Mountford, Agnes Cecile, Jeremy Mann, fluid oil and ink canvas, fine art, super dramatic light, photoillustration, breathtaking depth, hyperdetailed, uncanny, grim, creepy, iridescent red elements, masterwork by head of prompt engineering, cinematic
After the next election ...
OH SHIT
I WENT TOO OVERBOARD
"white house american government covered in blood half whitehouse is womens face tainted with blood and suffering government built on the blood of the people. WOMEN CRYING NEXT TO THE BUILDING DEATH DISPAIR FUCKING MALEVOLENT FUCKING EVIL FUCKING INSIDIOUS DICTATORSHIP!. M"
This one is crazy shit ...
Spider-Tick π
with cn?
ControlNet I guess ....
yep
why i get this)
There is a spider sitting on the face of a fat woman, a face of horror
not spiderman lol
what i do wrong every time)
With comfyui you can get some pretty disturbing stuff, those nodes are powerful
After election party ...
comfy its like programing language for promt yes
visual promt programming
clownshark where my images what i request?
lol translate google bad for ai
Good nite!
What am I looking at π
π
@clever oar
ive bought out the training deadass π
π
BRO AINT NO WAY
AHAHAHAA
I have come just in time 
I mean it's apparent to me that I've made a few acquaintances in here π
bro the fingers are almost perfect
yeah ahahaa
egirl training data is wild π
deadass
π
this is unhinged bro
ok lets get the other training data in there
this training data is fucking depressing bro
π
wild shit
You know, I am now thinking, what must be convenient about ComfyUI is that you can save model-specific workflows so you don't have to set everything up every single time like in A1111 when you switch π€
comfui was worse then automatic111
when i tried it
i use automatic11111
Integer π
lol
i asked chat gpt
program lang
to compare the pain
between hello kitty women and a man like david goggins
it said david goggins has more pain π
π
π
i thought the ai is woke lol
it actually got that right
chatgpt not too censored
chatgpt x this model is crazy
getting dall e level results
"In the depths of his soul, a tempest raged, unseen yet all-consuming, tearing at the very fabric of his being. He wore his sadness like a cloak of lead, its weight dragging him down into the abyss of despair. But unlike a tangible burden, his sorrow manifested as a silent tempest, raging within the confines of his own mind.
It was as if his heart had become a battlefield, each beat a struggle against the encroaching darkness that threatened to engulf him. Like a relentless adversary, his sadness gnawed at the edges of his sanity, leaving behind a trail of devastation in its wake.
At times, it felt as though he were drowning in an ocean of emotions, the waves of despair crashing over him with relentless fury. Each breath he took was suffused with a suffocating heaviness, as if the very air had become thick with the weight of his sorrow.
But perhaps the most insidious aspect of his sadness was its ability to erode his sense of self, leaving behind nothing but a hollow shell of the man he once was. It whispered lies of unworthiness and self-doubt, its venomous words seeping into the very core of his being like a poison.
In the darkness of night, when the world lay still and silent, his sadness became his only companion, its icy grip tightening around his heart with each passing moment. It whispered of lost dreams and shattered hopes, of paths not taken and regrets that lingered like shadows in the recesses of his mind.
Yet amidst the chaos of his inner turmoil, there flickered a tiny ember of hope, a glimmer of light amidst the darkness. For even in his darkest moments, he knew that his sadness was not the sum total of his existence, but merely a chapter in the ever-unfolding story of his life."
it accurately represents when youre overloaded with emotions tb
yes
of course bro
if its you yes
π₯
HAHAHAHAA
AHAAHAHAHHAAHAHA
thats crazy bro
grus mom
π
π
its bad control
as it was
this is probably the best one
"how did your heart mend so easy mine still bleeds"
image heart ripped open
not fixed
bleeding
i will try in my horror model )
lol
yk my college have art department
the aritsts there are so big egoed
chemical
this good
everytime im ready to make a change you turn around and fuck out all my brains
A man farts out a giant oak tree, this tree has faces coming out of the bark, the tree is mad, the man is happy, there is a cat
Enjoy
Hey, don't look at me, you're stuck again
Gotta shoot at me
Roll up and pretend you're a rapper!
I don't intend to waste my time on
To kick you off
Damn you decided to pester me?
No, no, don't be annoying, I came here to hang
Maybe we hung out
If I didn't know you were a rapper
Puff daddy, puff daddy running from the law, puff daddy was found hiding here, he was not scared, he has nice shoes
AEEEEEEEE
a fat woman with a huge ass turned towards the viewer emits green gases
Giant mech spider, spider is holding knives, spider has the head of puff daddy
Go help dad π
Going to shop
π
The vram requirements for this model is fucking crazy
I'm using 10gbs gram
Vram
Which?
Proteus with a shit ton of Loras and training
Learning better upscaling than what Iβve been doing
π₯
While Open AI's DALL-E 2, Midjourney, and Stable Diffusion are generating incredible images, they consistently share a strange flaw
Read more: https://www.sciencefocus.com/future-technology/why-ai-generated-hands-are-the-stuff-of-nightmares-explained-by-a-scientist/
π
π
Lora gets rid of it 99% of the time
They're lucky most of us CBA to controlnet our shit ahahaha
What a minute
If we used sd3d
And did all images ai is trained on for hands
Then did a 360 image of hands
We would get non deformed hands
Yup
Lol
And a million years later
Yeah
Bro
I just realised
Zendaya is in a new movie where she kisses two guys
π
The issue with Hollywood is
They appeal to. Avery niche demographic
Plus isn't zendaya like 16 π
I donβt think she is.
But, they also get paid millions per film, so she can party how she wants
Fair enough
I wonder what's gonna happen to entertainment
When ai picks up
Everyone is so arrogant against ai lol
Especially artists
Pretty sure you would dance naked on an ice cream truck if you were offered 20 mil
I went to the art department
And I took pics of their art
Then used clip to get a prompt and recreated it
Nah they were pissd π
Hell, we have jokes about trying to take a sh*t under parachute on someoneβs car. Would do that for free!
Then they asked me "you can't do hands "
I did hands and they got soo pissed
Lol
I'm gonna be honest it didn't really matter
People hate what they donβt understand.
They should not be physical.
Wouldn't even call it physical
One of them pushed me and I didn't move an inch π
You WERE taking pics of there art thoβ¦
I felt bad for her
Donβt do that lol
It was on display
I asked a teacher it's allowed
They had to allow their art to go on display tbf

You can, doesnβt mean you should.
Yeah fair enough
Not scolding⦠just sayin
That partβ¦. Is comedy.
Then I moved back tbf
Shoulda recorded it
She had dyed blue hair
I didn't view her as a threat I just felt bad
She was probably epic
Cause idk maybe she was genuinely upset
Atta girl for swinging
She was angryβ¦
She shouldnβt have swung
Also, she should be fair warned that even if she copyrights her work. People are still going to steal it, and lawyers wont to much if anything
So better her heart gets a little broken now then later.
This is mine, own full copyright on it. You can play with it, but god help you if you try and sell it. Lol
Stolen more times than I can even admit.
Lawyers dont give a fuq
The GOOD news⦠is if you pay your money and get your work copyrighted⦠your lawyers are free
(If you have a case)
Iβve had to fight stupid stuff like this.
new IPadapter on comfyui is incredible
here's a quick example
this plus this
equals this
Union
and tons of other shit
you can connect faces with it
yeah
all kinds of shit
you do need comfyui for the fancy shit just fyi
it's not hard at all to set up, the workflows are available on the github
this plus this
equals these
repaint the cat?
yes
reason #1,059,723 to get comfy with comfy
i have a feeling this won't be available anytime soon on a1111, if ever
how u doing that
@jovial tiger i have a feeling you'll really like this
this is the new ipadapter from cubiq on comfyui
name?
there's no prompt except just stuff like high quality photograph, etc
why is everything for comfy ui
It's literally called ipadapter
because comfyui is the easiest place to implement new stuff and the place where you can do the most advanced shit by far
i hiiiighly recommend just giving comfyui a shot if you haven't cuz you don't even need to learn anything to use it
there's default workflows you can just load and type shit into, or load images into to mix
Iβm learning about βpipeβ stuff now which is neat
Lots of those workflows are not great⦠some are!!! Some are a rats nest of hell
yup some make it too complicated
here's my advice: don't worry about organization
either use a premade workflow and tweak it a bit, or use it as a reference and start from scratch
and just fling shit everywhere until it's too messy to understand then delete shit you don't need or start over
agonizing over the organization is how you get stuck, or at least how i do
Iβm just now understanding the basic grasp of all its connection points and why. Took me a week of doing just that
Was begging folks for workflows to see what they were doing, and studying thoseβ¦ but people think Iβm asking for hand outs lol
don't hesitate to ask
i'm happy to share what i have
i also leave workflows in all my images
What kinda stuff are you working on?
so you should be able to click in discord "open in broswer" then just drag em in
Every image I'm posting has the workflow I used in it.
Same
Some are complex, some are not.
End goal, animation. Current learning curve is learning the upscale, and basics of animate
I have some good workflows I can share for photo.
i made a cat dog raccoon duck once
which transformed from one to the other in an infinite loop
that's about it
need go to work after 4 hour and sleep but i generate images )
OH SHIT
WE MIGHT HAVE A GEMINI SITUATION
π
fuck it we ball
lets let the loras out
@viral valve that's what it looks like
its little test
combined three images to make this one
my gpu gonna explode
huh
where the heck you getting these images from
Clownshark need combine many images for get horror like i have from normal promt
most people here generate stuff on their computer, locally.
what?
I'm making them.
alright
The ComfyUI workflow is embedded in the images.
i get horror every time
you're not very familiar with my work are you lol
today?>
He's literally making them right now.
oh ok
If work would stop getting in the way of my SD learning skills
just quit
Heh, you would have a hard time getting your graphics cards if I quit
pretty soon all video and image evidence will be inadmissable in court so a life of crime and stable diffusion will be the next great lifestyle
oh, do you work at nvidia?
Itβs ok. I work for BestBuy
man, i tried to buy my 4090 from BB but they kept flagging me as a potential scammer or some shit
even though i own my house, have no apt address, etc
my name and face is on a public website for my employer etc
That β¦ was hell.
finally gave up and bought from B&H
I can tell you why lol
A new system was built to circumvent robo buying. It was not perfect, but it needed to be done
ah, cuz of the scalpers?
got one more for ya
I wish you could drop in video the same you can drop on an image
Lovely!
I need to start building a giant mechanical owl, with clock internals
How about a cat holding a human
We have time. I get off work in 2 hours so my generations are on pause till then
yeah. this new ipadapter is sensational. holy shit
Last night was me literally living in animate hell
You wanna have some real fun. Take a picture of a chair, and tell it to make a cat chair π
i'm not using conditioning in any way, any attention masking, timestepping, etc...
